With increasing wind installations, there is a need to analyse wind generation variability in detail. This paper applies the reanalysis approach for modelling the variability; however, with two important additions. Firstly, high-resolution microscale data is combined with mesoscale reanalysis time series to model local variability in wind. Secondly, as there are often missing technical parameters in large-scale wind power plant datasets, machine learning is used to estimate the missing values. It is shown that such detailed modelling leads to more accurate simulations than a baseline model when compared to historical data from multiple European countries. In addition, applicability of the methodology for analysing future scenarios with changing wind installations is demonstrated.
- Random forest